What are the top healthcare AI startups by revenue today?

The ranking of top startups in the healthcare AI market by revenue

Below is a table ranking the startups of this market by their current revenue scale. You can find our methodology at the end of this page.

If you want a deeper understanding of the market and its current dynamics, get our report on the healthcare AI market.

Is healthcare AI already a billion-dollar business?

Healthcare AI has crossed the billion-dollar annual revenue mark, and Tempus AI is the company that got there.

Tempus AI and Caris Life Sciences both reported substantial fiscal-year sales for 2025. Their main business is cancer diagnostics: doctors order tests, laboratories process samples, and customers pay for the results. OpenEvidence follows a different route, with a reported annualized revenue run-rate as of July 2026. That describes its latest sales pace, not revenue booked over a completed year.

The difference tells us a lot about where healthcare AI makes money today. Hospitals and insurers already have budgets for diagnostic tests. Tempus and Caris can use AI to improve work that buyers have paid for years. A doctor-facing assistant must establish a new habit and a reason to keep paying for it.

Healthcare AI has real customers now, although a big annual total says little about profit or how quickly the next generation of software can catch up.

Does Tempus AI still earn more than Caris Life Sciences?

Tempus AI remains ahead of Caris Life Sciences on healthcare AI revenue reported for the same completed fiscal year.

Tempus AI recorded $1.27B of fiscal-year revenue for 2025, compared with Caris Life Sciences' $812M for 2025. Both numbers came from company results, so the comparison is unusually clear. Tempus combines genomic tests with a data-and-applications business. Caris makes most of its money from molecular profiling that helps oncologists choose treatments.

Caris's February 2026 results put its annual clinical therapy-selection volume at approximately 199,300 cases, up 22% year over year. The company also pointed to reimbursement and demand for its profiling services. Tempus has another way to earn money: pharmaceutical companies and researchers can pay for data and applications alongside the testing business.

That extra line of business helps explain Tempus's position, but it would be wrong to dismiss Caris as a smaller software challenger. Both companies sell a service that sits inside real treatment decisions. For now, the finished financial statements give Tempus the lead. What happens next will depend on test demand, insurer payments and whether the data businesses continue to grow.

Does Tempus AI really make its money from AI software?

No, Tempus AI earns most of its healthcare AI revenue through diagnostic testing rather than software subscriptions alone.

Tempus's 2025 results separate $955.4 million in diagnostics revenue from $316.4 million in data-and-applications revenue. Those are two quite different products. One involves collecting and analyzing medical samples; the other includes information and tools that help medical and research teams work with clinical data. Both use software and AI, but a diagnostic test also needs a laboratory, staff and payment from a patient, provider or insurer.

Caris Life Sciences has a similar diagnostic core. Innovaccer is closer to the familiar enterprise-software model: it connects healthcare information and sells tools to health systems and payers. Its recurring-sales measure therefore tells us something different from the revenue in Tempus's annual statement.

This matters when people imagine that the biggest AI company must be the one with the most widely used chatbot. The large sales in healthcare AI currently come from products with an established clinical reason to buy them. The technology can be impressive, but a doctor still needs to order the test and someone still needs to pay for it.

Could OpenEvidence overtake Tempus AI on revenue?

For now, OpenEvidence cannot claim the healthcare AI revenue lead over Tempus AI, even with its rapid reported growth.

The Information reported an annualized revenue run-rate for OpenEvidence in July 2026. That figure projects a current pace across a year, whereas Tempus reports sales actually earned during its 2025 fiscal year. We should not use the two figures as if they were identical accounting measures.

OpenEvidence has a different way to reach customers. Doctors can look up clinical evidence during a normal day without buying a new scanner or sending a sample to a laboratory. That makes the product easier to try. It also raises a question that matters more than download counts: how much of the use turns into lasting paid relationships?

Tempus earns money each time its diagnostic and data customers buy work that fits established medical budgets. OpenEvidence can grow quickly without those physical steps, but it must keep clinicians coming back and establish dependable payment. We see strong reported commercial progress, not proof that its annual recognized revenue has overtaken the diagnostic leader.

Is OpenEvidence earning more than Innovaccer?

The latest reported annualized sales pace favors OpenEvidence over Innovaccer, but neither figure is completed-year healthcare AI revenue.

OpenEvidence's reported ~$300M run-rate dates to July 2026. Fortune reported that Innovaccer crossed $200M ARR in the same month. A run-rate extends recent sales into an annual pace; ARR annualizes recurring subscription commitments. The numbers tell us both businesses are selling at scale, but they do not tell us that either collected that amount during one fiscal year.

The buyers explain why the comparison is interesting. OpenEvidence puts medical information directly in front of clinicians. Innovaccer connects patient records and builds data workflows for health systems and insurers. Hospitals can find enterprise software slow to install, but switching it off later may also be difficult. A search tool can spread faster among doctors, yet daily usage and long-term contract value are separate questions.

Komodo Health had already disclosed >$200M ARR as of December 2023. That older annualized figure gives useful context for healthcare data platforms, but cannot establish their sales pace today. The recent OpenEvidence and Innovaccer disclosures show that clinical search and institutional data software are both commercial categories, even if they sell to customers in very different ways.

Are healthcare AI diagnostics making more money than clinical software?

Healthcare AI diagnostics have the strongest published full-year sales, while clinical software is growing through subscriptions and annualized contracts.

Tempus AI and Caris Life Sciences earn their biggest reported sums from testing and related clinical data. HeartFlow sells analysis for coronary CT scans. In software, Innovaccer and Komodo Health report ARR, while OpenEvidence reports a run-rate. Those differences matter: the diagnostic companies have completed-year revenue figures, and the software firms often disclose a measure of current subscriptions or sales pace.

One reason diagnostics do well is simple. A cardiologist or oncologist already has a test to order and a medical reason to order it. When coverage and reimbursement are in place, hospitals can repeat the purchase. Software has its own attraction: a hospital might pay for a service across thousands of employees without sending each request through a laboratory. But it also takes time to install and integrate new systems.

We can compare published figures without pretending they all measure the same thing.

The table below shows selected companies in the order of our revenue ranking, with the measurement beside each one.

Startup	Figure	Measure	Period
Tempus AI	$1.27B	Fiscal revenue	FY2025
Caris Life Sciences	$812M	Fiscal revenue	FY2025
Komodo Health	>$200M	ARR	Dec 2023
Innovaccer	$200M	ARR	Jul 2026
HeartFlow	$176M	Fiscal revenue	FY2025
Lunit	~$61.5M	Fiscal revenue	FY2025
Aidoc	>$60M	Run-rate	Jun 2025
Suki	~$20M	Annual revenue	FY2025

Are hospitals really paying for HeartFlow's cardiac AI?

HeartFlow shows that hospitals really do pay for healthcare AI when the product helps answer a specific medical question.

HeartFlow recorded $176M of fiscal-year revenue in 2025 in an SEC filing. HeartFlow's filings already show $190M for the first nine months of 2026, so the ranking uses its last completed year rather than its faster current pace. Its software analyzes coronary CT scans to help cardiologists assess blockages and blood flow. This is not a general assistant that a hospital is experimenting with. Doctors can request a defined analysis as part of an existing diagnostic decision.

The company reported 1,465 US accounts at the end of 2025. In January 2026, HeartFlow announced that Aetna would cover its plaque-analysis product, adding another insurer to its coverage arrangements. Coverage does not guarantee that a scan will be ordered. It does make the financial part of the decision easier for providers and patients.

HeartFlow has spent years working on clinical studies, regulatory requirements and payer access. That is a slower way to grow than offering a chatbot online, but it creates a reason for doctors to use the product repeatedly. We see a practical lesson for other clinical AI startups: convincing a hospital that the tool works is only one step. Getting it into a paid care pathway is where sales start.

Can AI drug discovery startups earn money before selling drugs?

Drug discovery startups already earn revenue from scientific collaborations, long before their AI-designed medicines reach patients.

XtalPi reported ~$75.3M of 2025 revenue from its drug discovery solutions segment, the only part of the company we count in healthcare AI. Insilico Medicine reported $56.2M of fiscal-year revenue for 2025, then $106M for the first half of 2026 alone, though the ranking keeps its completed year. Generate:Biomedicines recorded $31.9M of collaboration revenue for that year, while Iambic Therapeutics reported $12.8M for the first half of 2026. The last two figures cover a narrower activity or period, so they cannot be treated as full-year equivalents.

A pharmaceutical company can pay for discovery work, a license or research milestones while a treatment is still in development. That money can become recognized revenue years before a patient buys a drug. Insilico's disclosed sales, for instance, do not mean an AI-discovered treatment produced the same amount at pharmacies. Success in a clinical trial remains a separate question.

XtalPi's mix of scientific tools and automation offers another route to customers. It can sell work and equipment linked to drug development without relying on one molecule reaching the market. These businesses therefore have real paying partners, but their revenue may jump when a large collaboration is signed and slow down between deals. We would be careful about reading a smooth subscription business into their results.

Is medical imaging AI getting paid, or just getting FDA clearance?

Medical imaging AI is making real sales, even though FDA clearance alone says nothing about how much a hospital will buy.

The FDA says it had authorized more than 1,600 AI-enabled medical devices by September 2026. That count includes devices across medicine and is not a count of revenue-generating startups. For the sales side, Lunit reports filed annual revenue, Aidoc has an annualized run-rate disclosure, and Qure.ai recorded ~$21.1M of fiscal-year revenue for the year ended March 2025.

Qure.ai's January 2026 impact report describes deployments across more than 4,500 healthcare facilities in over 100 countries. Its tools include tuberculosis screening. Aidoc concentrates on detecting urgent findings and getting them to clinical teams. Lunit works in cancer screening and other imaging applications. Different products can be cleared for use but face very different hospital budgets and workflows.

A hospital must integrate the software, train people, decide who pays and keep using it after the first rollout. Those steps explain why medical clearance and revenue do not rise together automatically. We look for documented sales alongside clinical adoption, rather than treating another FDA listing as proof of a successful business.

Is Lunit actually selling more medical AI than Aidoc?

Lunit has a slightly higher displayed figure than Aidoc, but their healthcare AI disclosures cannot prove which one earns more annual revenue today.

Lunit filed ~$61.5M of fiscal-year revenue for 2025 in South Korea. An investment fund reported >$60M of annualized revenue run-rate for Aidoc as of June 2025. The figures are close, but Lunit's amount covers a completed year while Aidoc's is a minimum annualized pace. We keep their established order without claiming that one reported more like-for-like annual sales.

Lunit sells imaging products for cancer screening and other oncology work. Aidoc focuses on urgent radiology findings and helps medical teams act on them faster. Both have regulated products, but hospitals may buy them for different parts of the clinical day.

For investors and customers, the useful question is whether clinicians keep opening the software after it is installed. A new regulatory clearance can add a product to the catalog. Regular use and contract renewals are what turn it into sales. Lunit's filed annual revenue offers firmer evidence of that scale; Aidoc's run-rate suggests a substantial current pace with a different reporting standard.

Are hospitals paying real money for AI medical scribes?

Yes, AI medical scribes have become paid healthcare software businesses, with Heidi Health and Suki offering useful evidence.

Heidi Health has disclosed ARR as of April 2026. Suki reported ~$20M of annual revenue for 2025 through press coverage, while Freed has a company-disclosed $10M ARR figure with no stated period. ARR annualizes recurring subscriptions, so it is not identical to a company's recognized sales for a completed year.

The job is easy to understand. A clinician talks to a patient, and the software helps prepare the medical note. Heidi's September 2026 announcement said its products were involved in millions of patient visits each week. It has since moved beyond transcription toward tools that assist with more steps of the care process.

That expansion could make these products more useful to hospitals, but it also brings more responsibility. A bad summary wastes a doctor's time; an incorrect next step in patient care can do real harm. We would put more weight on paid renewals and careful deployment than on the number of notes generated. The established sales figures already show that documentation saves enough work for customers to pay for it.

Is Abridge making more money than Heidi Health?

There is no verified healthcare AI revenue comparison that puts Abridge ahead of Heidi Health today.

Heidi Health disclosed $50M ARR as of April 2026. Sacra estimated $100M ARR for Abridge as of May 2025, but that is an independent analyst estimate, not a company-reported financial statement. With no second independent confirmation, Abridge has no revenue rank in this comparison. Treating the estimate as confirmed annual sales would give readers more certainty than the evidence allows.

Abridge is well known for its hospital documentation deployments. Those partnerships matter because a deal with a health system can reach many clinicians. The contract headline, however, does not tell us how many subscriptions have started or how payments are recorded. Ambience Healthcare faces a similar problem when outsiders try to work out its financial scale from adoption announcements.

We can say Heidi has publicly disclosed a substantial recurring-revenue business. We cannot say Abridge earns less just because its figures come from outside analysts. That distinction leaves room for Abridge to be larger without awarding it a position that its verified financial disclosures do not support.

Are healthcare AI agents making money beyond medical notes?

Healthcare AI agents already earn money from scheduling, referrals and medication access alongside medical documentation.

Forus disclosed >$10M of annualized revenue run-rate as of December 2025, a sales pace at a point in time rather than a finished year of recognized revenue. Assort Health has no revenue rank: estimates from CB Insights and Forbes put its run-rate anywhere from $3M to $30M as of June 2026, too wide a range to compare. Assort helps practices handle appointment and referral work; Forus focuses on getting patients access to prescribed medicines.

Assort announced referral-handling agents in August 2026 and later added digital intake and payments. The company is trying to handle more of the jobs that come between a patient contacting a practice and arriving for care. The potential customer benefit is concrete: fewer repeated calls, faster booking and less staff time spent copying information between systems.

Healthcare administration is full of exceptions. A referral may be urgent, incomplete or sent to the wrong clinic. That is why reliable handling matters more than a smooth voice demo. Forus's reported run-rate shows customers are paying for automation now, but it does not prove these tools can safely make every decision without staff supervision.

Can healthcare AI startups outside the US compete on revenue?

Outside the United States, healthcare AI startups have substantial reported sales, though US diagnostic businesses remain ahead on published fiscal-year revenue.

China's XtalPi sells scientific technology and laboratory automation. South Korea's Lunit sells imaging AI, while Israel's Aidoc reports an annualized pace from radiology tools. Australia's Heidi Health sells clinical documentation software. Germany's Ada Health, Britain's Huma and India's Qure.ai also have financial disclosures. These are paying businesses in several national healthcare systems.

The route to a contract changes from country to country. Some companies sell through hospitals and national health systems. Others work with pharmaceutical developers, diagnostic labs or private practices. We cannot declare a regional winner by mixing a 2024 fiscal-year statement with a 2026 ARR disclosure, but it is already clear that healthcare AI revenue is not confined to Silicon Valley.

The table below shows selected non-US companies, using their published figures and the order of the established revenue ranking.

Startup	Country	Figure	Measure
XtalPi	China	~$75.3M	FY2025 segment revenue
Lunit	South Korea	~$61.5M	FY2025 revenue
Aidoc	Israel	>$60M	Jun 2025 run-rate
Insilico Medicine	Hong Kong	$56.2M	FY2025 revenue
Heidi Health	Australia	$50M	Apr 2026 ARR
Ada Health	Germany	~$40.7M	FY2024 revenue
Huma	UK	~$38.5M	FY2024 revenue
Qure.ai	India	~$21.1M	FY2025 revenue
Gleamer	France	$10.4M	FY2025 revenue

Is cancer testing where healthcare AI makes the most money?

Cancer testing is behind the largest published healthcare AI revenue figures, especially those of Tempus AI and Caris Life Sciences.

Both companies sell molecular diagnostics that help doctors choose cancer treatment. Lunit sells oncology imaging tools, and Freenome reported $30.4M of fiscal-year revenue for 2025 in blood-based cancer screening. A later Freenome filing shows $190M for the nine months to September 2026, but the ranking keeps its completed 2025 year. Their products and stages of commercialization differ, but they share a medical question customers already spend money to answer: which test can help decide what to do next?

Caris's February results showed rising numbers of therapy-selection tests. Tempus can also earn money by licensing clinical data to researchers and pharmaceutical companies. Patients may need testing, monitoring and follow-up as treatment changes. This creates repeated reasons to buy the underlying services.

Still, a company building an experimental cancer-detection model is not automatically going to match a diagnostic laboratory's revenue. Approval, insurance payments and routine clinical use all take work. The opportunity is real because cancer care has established buyers, not because every model aimed at oncology already has a paying market.

Are European radiology AI startups actually making money?

Yes, European radiology AI startups have recorded sales, and some have already attracted major medical-technology buyers.

RadNet's DeepHealth acquired France's Gleamer in March 2026. Gleamer reported $10.4M of fiscal-year revenue for 2025 in French accounts. Britain's Brainomix recorded ~$7.77M of fiscal-year revenue for 2025, and France's Incepto Medical disclosed ~$7.59M of segment revenue for that year. Incepto's amount covers one segment, so it cannot be compared as a complete company total.

RadNet said Gleamer had more than 700 customer contracts across 44 countries when it announced the purchase. That helps explain the deal: Gleamer already had hospital relationships and imaging tools that fit routine radiology. Brainomix concentrates on stroke and lung scans. Incepto sells diagnostic applications through clinical imaging channels.

European startups have found real customers, but a hospital installation is only the start. Hospitals still need to use the software regularly and renew their contracts. We can point to filed revenue and an acquisition as evidence of a business. We cannot assume every company with a cleared product has achieved the same scale.

Why are big medical companies buying healthcare AI startups?

Large medical companies are buying healthcare AI startups to add useful software to the clinical products and sales channels they already own.

Agilent Technologies completed its purchase of Proscia on November 4, 2026. Proscia had disclosed >$18M of trailing-twelve-month revenue as of October 2026. Tempus AI bought Paige on August 22, 2025; its filing records $4.6M of segment revenue for Paige's post-acquisition months in 2025. RadNet's DeepHealth acquired Gleamer in March 2026, adding routine radiology tools and hospital customers.

The products solve different problems. Proscia helps labs work with digital pathology images, Paige brings AI tools for cancer slide analysis, and Gleamer supports everyday imaging. An acquirer can connect those tools to instruments, laboratories, hospital contracts and existing sales staff. That may be faster than building a competing product from scratch.

The revenue figures describe different periods, including one acquired company's partial-year results. They do not prove that the buyers will earn back their purchase prices. But the deals show something important about today's market: medical AI software is valuable enough for established healthcare businesses to put it inside their own offerings.

Can a billion-dollar healthcare AI contract count as revenue?

A billion-dollar healthcare AI deal can be genuine without creating a billion dollars of earned revenue.

Earendil Labs disclosed >$1.5B of potential contract value for a drug-discovery collaboration. Profluent announced a $2.25B contract-value agreement in April 2026. Such agreements can include payments spread over years and milestones that may never be reached. They should not be confused with annual sales.

Commure reported >$25B of annualized GMV from claims and payments processed as of January 2026. That is the value of activity moving through the platform, much of it belonging to providers and insurers. Isomorphic Labs described $82.5M in upfront collaboration payments, another financial measure that does not establish comparable annual revenue. Tandem Health's 10K care organizations and Sirona Medical's >12M imaging exams a year under contract show customer reach rather than dollars earned.

These announcements tell us which companies have partnerships, distribution or commercial demand. The amounts can be impressive. We still need an accounting period and a revenue measure before they can answer who makes the most money. Without that distinction, a multiyear potential contract could appear bigger than years of actual sales.

Are the biggest healthcare AI startups profitable yet?

Even the highest-revenue healthcare AI startups can lose money, and several do.

Tempus AI reported a net loss in its full-year 2025 results. Caris Life Sciences also reported a net loss for 2025 despite increasing clinical test volume and sales. Lunit's financial statements likewise distinguish revenue from operating results. The companies can be commercially successful at selling services while still spending more than they earn overall.

Their costs also look different. Diagnostic companies run laboratories, handle medical samples and work with insurers. Imaging software vendors pay for clinical studies, regulatory approvals and hospital integration. AI scribe companies avoid the lab but have computing, compliance, support and implementation bills. A high revenue figure tells us how much a product sells, not how much cash it keeps.

This is especially important for a sector where many companies reveal an ARR number but little detail about expenses. Revenue gets a startup into a serious commercial discussion. Profit needs its own evidence. We would not declare a company profitable from its sales figures alone.

Are famous healthcare AI startups actually earning real revenue?

Some famous healthcare AI startups may be earning substantial revenue, but their public numbers are too thin to compare confidently.

Abridge has a $100M ARR analyst estimate for May 2025, not a company-verified figure. Cohere Health's $100M annual revenue estimate for 2024 comes from an automated company-data service. Candid Health has a $57M annual revenue estimate without a stated period. The figures are worth knowing, yet none comes with the independent confirmation needed for a defensible revenue rank.

Hippocratic AI, Notable and other frequently discussed names face a similar disclosure problem. Some companies talk publicly about patients, customers or the work completed by their software. Infinitus cited >3.7M patients supported, a number that does not say how much customers paid. Neko Health has only a $21.9M automated estimate of annual revenue for the year to September 2025, with no second independent source, so it is unranked. Assort Health's estimates range from $3M to $30M, which leaves it unranked too. Function Health's separate $100M analyst estimate of annualized run-rate as of February 2025 remains unranked.

That leaves a gap between publicity and measurable sales. We would rather say that a company might be larger than its public figures suggest than fill the gap with invented revenue. A well-known name can be a serious business without having published a comparable annual statement.

So, which healthcare AI startups make the most money today?

Tempus AI is the number one healthcare AI startup by revenue today, based on its published fiscal-year results.

Caris Life Sciences comes next on completed fiscal-year revenue, followed in the established ranking by OpenEvidence's annualized run-rate, Komodo Health's older ARR and Innovaccer's newer ARR. Those figures answer slightly different questions about commercial scale. The cleanest comparison at the top remains the two large diagnostic businesses with completed-year statements.

There is plenty of paid activity below them. HeartFlow has a reimbursed cardiac product. XtalPi and Insilico Medicine earn money from research, technology and drug-discovery work. Heidi Health and Suki sell software that removes real tasks from clinicians, while Forus automates work for practice staff. Lunit and Qure.ai show how imaging AI can turn deployments into reported sales across borders.

We see several working business models rather than one winning form of healthcare AI. For now, cancer diagnostics lead on published annual revenue. Clinical tools and administrative agents may catch up, but their annualized measures cannot be quietly turned into fiscal-year sales. Famous startups with only forecasts, bookings or outside estimates remain impossible to place with the same confidence.

The table below shows what we found.

Question	Answer
Do cancer tests still beat chatbot subscriptions?	Yes, diagnostic testing tops the published healthcare AI revenue figures. Tempus recorded $955.4 million from diagnostics in 2025. Doctors already have a clinical reason to order those tests.
Could Caris be dismissed as a minor challenger?	No, Caris is a sizable cancer diagnostics business. It completed approximately 199,300 clinical therapy-selection cases during 2025. Those tests show real demand behind the company's financial results.
Has doctor-facing AI search become a business?	Yes, OpenEvidence has reported a substantial annualized sales pace. Its run-rate was ~$300M as of July 2026. It still cannot replace a completed fiscal-year revenue statement.
Do older subscription figures prove current growth?	No, a historical ARR disclosure shows past commercial scale. Komodo Health reported >$200M ARR as of December 2023. Today's sales could differ, so the date really matters.
Can hospital data platforms sell big subscriptions?	Yes, Innovaccer sells software to health systems and payers. It reported $200M ARR as of July 2026. That figure says nothing about its profit after expenses.
Can hospitals pay repeatedly for cardiac AI?	Yes, HeartFlow sells analysis used in clinical decisions. It reported $176M of fiscal-year revenue for 2025. Insurance coverage helps make that service a routine purchase.
Is AI drug discovery already commercially useful?	Yes, XtalPi sells drug discovery solutions to drug developers. Its drug discovery solutions segment earned ~$75.3M in 2025. Customers can pay for discovery work before drugs reach patients.
Are AI medical scribes generating subscription sales?	Yes, Heidi Health has an established clinical documentation business. The company disclosed $50M ARR as of April 2026. That amount annualizes subscriptions rather than completed-year recognized revenue.
Do patient-access agents generate paying customers yet?	Yes, Forus sells automation that helps patients get prescribed medicines. Its run-rate was >$10M as of December 2025. Practices pay to remove repetitive administration from their days.
Have independent imaging models moved beyond pilots?	Yes, Qure.ai has filed imaging AI revenue figures. Its fiscal-year revenue was ~$21.1M through March 2025. Medical deployments need actual sales to prove business demand.
Does a billion-dollar partnership mean billion-dollar sales?	No, long-term contract value is not earned annual revenue. Profluent announced $2.25B of potential contract value in April 2026. Payment depends on future work and agreed development milestones.